|
[DPRG] offroad robot exercises
Subject: [DPRG] offroad robot exercises
From: David P. Anderson
dpa at io.isem.smu.edu
Date: Thu Oct 26 14:08:34 CDT 2006
Howdy
I've been thinking about our outdoor robots and what sorts of things we
might do at the November DPRG meeting to help everyone improve their
robot's performance.
I have a couple of suggested exercises that have been helpful to me in
developing jBot. These are exercises we can run at the Dallas Fairgrounds
Saturday afternoon after the DPRG meeting at The Science Place, if anyone
is interested.
Building a robot requires getting a lot of very basic hardware working including
(but not limited to) attaching wheels and motors and gearheads and maybe shaft
encoders to a platform, some sort of H-Bridge(s) to control the motors, battery
power for the motors and electronics, some sort of computer or micro-controller
with its attendant development environment all set up and working, as well as
developing software to generate control and PWM signals and read shaft encoders
and other inputs, and drive the robot. All of this must be robust enough to
handle being offroad, and allow development of higher-level robot behaviors
without having to constantly re-visit electrical and mechnical problems.
So a realistic FIRST exercise for folks attempting to build an offroad robot is to:
1. Drive in a straight line to a distant point and back, with no obstacles.
100 feet is a good test. (1000 would be better!)
This will test all of the basic hardware and software robot systems without
getting into more complex navigation and avoidance behaviors.
Driving in a straight line between two points seems like a trivial task, and
perhaps it should be. Nonetheless, lots of robots get hung up at this stage of
development, and never make it any further. It's a good bet that, if your robot
can't do this simple exercise, it will have trouble with more difficult navigation
tasks. So this is a good first challenge to shoot for in building an offroad robot.
As a good reality check, look through the videos of the SRS Robomagellan that
Dave Hylands has so graciously made available. I believe that contest is into
its 3rd year, but many experienced robot builders are still having problems at
this basic level.
<http://davehylands.com/Robotics/Robo-Magellan/>
The SECOND offroad robot exercise is only a slight variation of the first:
2. Drive around a large square, clockwise or counter clockwise, and stop at
the starting point, with no obstacles. 100 feet is a good size (400 feet total).
Now that we have our robot working well enough to do exercise #1, exercise #2 can
be used to test and also to actually calibrate the location calculations of the robot,
as J.Borenstein describes in his UMBMark paper:
<http://www-personal.engin.umich.edu/~johannb/Papers/umbmark.pdf>
Driving around a square, rather than just a straight line, is a better test of
the robot's ability to keep track of its location. As in exercise #1, it doesn't
really matter if the robot goes exactly to each waypoint, but only its final
stopping place, which is presumably an accumulation of all errors up to that time.
By measuring the distance between where the robot stops, (i.e., where the robot
"thinks" the starting point is) and the actual starting point, we get an error
in X and Y that can be used to calibrate the location calculations.
For example, using these measurements for 100 foot squares driven clockwise and
counterclockwise, I was able to determine that the magnet compass on jBot was
being influenced by the magnetic fields of the robot's electric motors. So the
senor was mounted on a mast, and the mast was made taller and taller, experimentally,
until that error went away when driving the Borenstein squares.
There are several places at Fairpark, just outside the Science Place, where we
can run these sorts of exercises. We can either do it as a fixed distances in some
arbitrary direction, or I also have some GPS waypoints I'll post separately for
those who'd rather use lat/lon coordinates.
The third exercise is for those who have obstacle avoidance working and want to
test that out, and it is very simple.
3. Drive exercise #1, out and back, but with obstacles in the way.
A good example might be a small grove of trees between the two waypoints, or a
fence or building, or even the corner of a building, like this:
B
|
|
tree tree
tree tree
tree tree
tree
tree tree tree
tree tree
|
|
A
B
| /
| /
| /
|/
Building |
|
|
-----------------
/
/
/
/
A
Both of those scenarios are availble in the meadow just east of the Science Place.
I'm not sure anyone will be ready for obstacle avoidance by next month, but if
we can at least get several robots to run exercise #1 and #2, we will be well on
our way to the next navigation contest.
best regards,
dpa
More information about the DPRG mailing list
|
